dfs health Dashboard

[V Satish Kumar <sa...@mkhoj.com>]

Hi,

I have noticed that the dfs health dashboard(Running on port  50070) 
takes a long time to refresh the number of live nodes and dead nodes. Is 
there a config parameter in hadoop that can be changed to make the 
dashboard shows these changes more quickly?

Thanks,
Satish

Re: dfs health Dashboard

[Koji Noguchi <kn...@yahoo-inc.com>]

Hi Satish, 

This doesn't solve your current problem, but from 0.20 (after HADOOP-4029),
"4. List of live/dead nodes is moved to separate page. "

Koji


On 11/1/09 11:11 PM, "V Satish Kumar" <sa...@mkhoj.com> wrote:

> Hi,
> 
> I have noticed that the dfs health dashboard(Running on port  50070)
> takes a long time to refresh the number of live nodes and dead nodes. Is
> there a config parameter in hadoop that can be changed to make the
> dashboard shows these changes more quickly?
> 
> Thanks,
> Satish

Re: dfs health Dashboard

[V Satish Kumar <sa...@mkhoj.com>]

Hi Allen,

 Once a node goes down, the dfs health dashboard takes 40 to 45 minutes 
to refresh the status from live to dead.
 But I see that the 'Last Contact'  field at Live 'dfsnodelist' page 
getting updated every 30 seconds.





Allen Wittenauer wrote:
>
> On 11/1/09 10:11 PM, "V Satish Kumar" <sa...@mkhoj.com> wrote:
>   
>> I have noticed that the dfs health dashboard(Running on port  50070)
>> takes a long time to refresh the number of live nodes and dead nodes. Is
>> there a config parameter in hadoop that can be changed to make the
>> dashboard shows these changes more quickly?
>>     
>
> What do you mean by "long time"?  Hours? Days?
>
> Keep in mind that you don't want a failure to be "instantaneous" because
> that means every time there is even a minor, temporary blip the namenode and
> the network will be busy replicating blocks that are just going to re-appear
> in a minute or so anyway.
>
> .
>
>   

Re: dfs health Dashboard

[Allen Wittenauer <aw...@linkedin.com>]

On 11/1/09 10:11 PM, "V Satish Kumar" <sa...@mkhoj.com> wrote:
> I have noticed that the dfs health dashboard(Running on port  50070)
> takes a long time to refresh the number of live nodes and dead nodes. Is
> there a config parameter in hadoop that can be changed to make the
> dashboard shows these changes more quickly?

What do you mean by "long time"?  Hours? Days?

Keep in mind that you don't want a failure to be "instantaneous" because
that means every time there is even a minor, temporary blip the namenode and
the network will be busy replicating blocks that are just going to re-appear
in a minute or so anyway.